Dimensionality Reduction for Supervised Learning in Link Prediction Problems

Abstract

In recent years, a considerable amount of attention has been devoted to research on complex networks and their properties. Collaborative environments, social networks and recommender systems are popular examples of complex networks that emerged recently and are object of interest in academy and industry. Many studies model complex networks as graphs and tackle the link prediction problem, one major open question in network evolution. It consists in predicting the likelihood of an association between two not interconnected nodes in a graph to appear. One of the approaches to such problem is based on binary classification supervised learning. Although the curse of dimensionality is a historical obstacle in machine learning, little effort has been applied to deal with it in the link prediction scenario. So, this paper evaluates the effects of dimensionality reduction as a preprocessing stage to the binary classifier construction in link prediction applications. Two dimensionality reduction strategies are experimented: Principal Component Analysis (PCA) and Forward Feature Selection (FFS). The results of experiments with three different datasets and four traditional machine learning algorithms show that dimensionality reduction with PCA and FFS can improve model precision in this kind of problem.

The use of machine learning algorithms in recommender systems: A systematic review

“…Liu et al, 2011;Q. Liu et al, 2014;Lops et al, 2009;Luong, Huynh, Gauch, Do, & Hoang, 2012;Marović et al, 2011;Marques et al, 2011;McLaughlin & Herlocker, 2004;Montañés et al, 2009;Moreno et al, 2012;Murfi & Obermayer, 2009;Musto et al, 2010;Nicol et al, 2014;Nie et al, 2013;Oyama et al, 2012;Pantraki & Kotropoulos, 2015;Paparrizos et al, 2011;Pecli et al, 2015;Pessiot et al, 2007;Pronoza et al, 2016;Roh et al, 2003;Y. Song et al, 2011;I.…”

“…Song et al, 2011;Xuan et al, 2014;Zahra et al, (Aouay et al, 2014;Bar et al, 2013;Biancalana et al, 2011;Braida et al, 2015;Buabin, 2012;Elmongui et al, 2015;T.-K. Fan & Chang, 2010;Islam et al, 2015;Kao & Fahn, 2013;Middleton et al, 2004;Szymański & Rzeniewicz, 2016;Tsuji et al, 2014;Vialardi et al, 2011;Yan et (Agarwal, 2011;Banerjee et al, 2012;Brouard & Pomot, 2016;Diaby et al, 2013Diaby et al, , 2014Ghazarian & Nematbakhsh, 2015;Jun, 2005;Kong et al, 2013;Pecli et al, 2015;Pronoza et al, 2016;I. Song et al, 2011;Sun et al, 2015;Szymański & Rzeniewicz, 2016;Tsuji et al, 2014;Verma et al, 2016;Yap et al, 2005 ACCEPTED MANUSCRIPT…”

“…Song et al, 2011;Verma et al, 2016;Wan et al, 2009;Yan et al, 2013;Yuan et al, 2014;K. Zhao & Pan, 2015) Academic 12 (Das et al, 2013;Hernández del Olmo et al, 2009;Huang & Nikulin, 2014;Krohn-Grimberghe et al, 2011;Luong, Huynh, Gauch, Do, & Hoang, 2012;Middleton et al, 2004;Montañés et al, 2009;Oyama et al, 2012;Pecli et al, 2015;Vialardi et al, 2011;Xin et al, 2014) News 11 (Bellogín et al, 2011;Brouard & Pomot, 2016;Buabin, 2012;Das et al, 2013;Z. Fan et al, 2016; W.-P. Lee & Lu, 2003;Leopairote et al, 2013;Q.-C. Li et al, 2008;Nguyen et al, 2016;Nie et al, 2013;Nicol et al, 2014) E-commerce 10 ( Anaissi & Goyal, 2015;Bauer & Nanopoulos, 2014;Buettner, 2016;Castro-Herrera et al, 2009;R.…”

“…Fan et al, 2016; W.-P. Lee & Lu, 2003;Leopairote et al, 2013;Q.-C. Li et al, 2008;Nguyen et al, 2016;Nie et al, 2013;Nicol et al, 2014) E-commerce 10 ( Anaissi & Goyal, 2015;Bauer & Nanopoulos, 2014;Buettner, 2016;Castro-Herrera et al, 2009;R. Zhang & Tran, 2010;Nie et al, 2013;Pecli et al, 2015;Wei et al, 2011;Q. Zhao et al, 2015;Zhai & Li, 2015) Webpages 10 (Forsati et al, 2009;Forsati & Meybodi, 2010;Jin et al, 2005;Jun, 2005;Kao & Fahn, 2013;J.…”

La identificación de enlaces ausentes como competición Kaggle para la enseñanza de teoría de redes

“…La tercera categoría es la de los métodos basados en la reducción de la dimensionalidad. En esta aproximación se formaliza el problema como de clasificación binaria, lo cual conlleva la conocida maldición de la dimensionalidad en aprendizaje supervisado (Pecli et al 2015); por este motivo, se emplean diferentes estrategias para reducirla. Por último, en el cuarto grupo se recogen otras aproximaciones basadas en estrategias diversas: aprendizaje, teoría de la información, métodos de perturbación estructural y aproximaciones mixtas, entre otras.…”